Author |
Message |
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2653 ID: 1 Credit: 109,635,410 RAC: 47,548
                     
|
During the 2019 BOINC workshop there was quite a lot of talk about people wanting to buy compute resources for their chosen projects, but not wanting to invest into hardware and take up the maintenance of it, for various reasons (difficulty setting up, high electricity costs, etc). Instead people would like to pay for cloud resources and adjust the donated amount as their income or engagement changes.
I went ahead and built the system for this: https://thescience.cloud/
It is very simple to start: just add money into your account, pick your projects and run as many computers as you wish. The computers are automatically started on a cloud provider and run for as long as you have money in your account. The computers get attached to your account (you have to add your weak account key), so all the BOINC credits also go to you and your team.
Supported payments methods are PayPal, ACH transfer in the United States, or SEPA transfer in European Union Eurozone. I add a small 7% markup which will hopefully cover PayPal fees, accounting costs and my hosting needs. 21% Lithuanian VAT for European Union / European Economic Area customers is included, this is run through a legal company I operate. Worldwide users currently do not have to pay tax based on Lithuanian law.
Special thanks to the initial testers from PrimeGrid's Discord! |
|
|
|
What would be nice is maybe a table of pricing before creating an account. I wanted to see what I might be able to do in the future when I hopefully get life back on a more steady basis, but I didn't want to make an immediate dormant account.
____________
My lucky numbers are 121*2^4553899-1 and 3756801695685*2^666669±1
My movie https://vimeo.com/manage/videos/502242 |
|
|
KEP Send message
Joined: 10 Aug 05 Posts: 303 ID: 110 Credit: 13,001,669 RAC: 43,318
          
|
Very nice Rytis :)
Just what kind of computers (CPUs) are we getting for 7 USD a month?
As far as I can see, I can actually rent 7 computers for the cost that I currently have to run just 1 computer (very nice) :) But how do they compare to an i5-4670? |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2653 ID: 1 Credit: 109,635,410 RAC: 47,548
                     
|
What would be nice is maybe a table of pricing before creating an account.
Will add that, thanks for the comment! |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2653 ID: 1 Credit: 109,635,410 RAC: 47,548
                     
|
Just what kind of computers (CPUs) are we getting for 7 USD a month?
For $7/mo you get one core on a Xeon E5-2650 v4, Xeon Gold 6140 CPU or similar. Note that it is likely to change as the current cloud provider is not very happy about the load we're causing on the servers - if we need to move, the offered CPUs and the price may change. |
|
|
Ken_g6 Volunteer developer
 Send message
Joined: 4 Jul 06 Posts: 940 ID: 3110 Credit: 265,153,553 RAC: 110,745
                            
|
So, from my perspective as someone who doesn't pay VAT, this is almost a 30% markup mainly for having BOINC set up for us? I'll go ahead and set up BOINC myself, thanks.
But I do hope you'll let us know if you find a reasonably-priced cloud provider that is happy about the load PrimeGrid can put on it. |
|
|
|
A dedicated CPU option is available with the cloud providers but at a higher cost than the shared CPU options.
DigitalOcean has the $5/month (plus tax) for the shared CPU. For a 'CPU-Optimized' service then it is $40 for a 2vCPU option.
That is $20 per core compared to $5 per core (memory is higher so that factors in to higher price however for our purposes here, it is irrelevant as we are concerned with computing power).
Azure had similar price differential however it was higher than DigitalOcean and varied based on location of server.
I feel that you need more information on your webpage as to the CPU's that you will be running the BOINC instance on as it affects how the projects will run. The two CPU's that you have mentioned have different capabilities where one supports AVX2 and the other AVX512. This is one of the few BOINC projects where that will have much significance but it is relevant.
I can see good potential for you if you market it properly. You need to target those who don't know how to set up their own cloud servers or those who don't have the time to set them up yet want to contribute. |
|
|
Ken_g6 Volunteer developer
 Send message
Joined: 4 Jul 06 Posts: 940 ID: 3110 Credit: 265,153,553 RAC: 110,745
                            
|
I feel that you need more information on your webpage as to the CPU's that you will be running the BOINC instance on as it affects how the projects will run. The two CPU's that you have mentioned have different capabilities where one supports AVX2 and the other AVX512. This is one of the few BOINC projects where that will have much significance but it is relevant.
I can see good potential for you if you market it properly. You need to target those who don't know how to set up their own cloud servers or those who don't have the time to set them up yet want to contribute.
True. Also, you should share RAM sizes. PrimeGrid doesn't much care about RAM, but some other projects do. NFS@Home is one, which happens to be concerned with prime numbers. Certain World Community Grid projects also need lots of RAM. |
|
|
|
When I saw this post, I hoped it was the reverse of what it actually is. I was hoping it was a way for someone else to use my home computer to run their projects for a small fee (which would have to be less than cloud providers to not be exploited). As someone who lives in an area with very low electricity cost, this would be attractive to me and then I would feel less bad about leaving my computer on 24/7 crunching things. Of course, you would also have to have the fee be associated with families of processors, and have requirements around stolen CPU time, etc.
____________
|
|
|
|
Azure had similar price differential however it was higher than DigitalOcean and varied based on location of server.
I just checked and they are offering $200 free credit and which is good for 30 days. |
|
|
mikey Send message
Joined: 17 Mar 09 Posts: 1895 ID: 37043 Credit: 825,277,592 RAC: 580,819
                     
|
When I saw this post, I hoped it was the reverse of what it actually is. I was hoping it was a way for someone else to use my home computer to run their projects for a small fee (which would have to be less than cloud providers to not be exploited). As someone who lives in an area with very low electricity cost, this would be attractive to me and then I would feel less bad about leaving my computer on 24/7 crunching things. Of course, you would also have to have the fee be associated with families of processors, and have requirements around stolen CPU time, etc.
You could do this on your own thru PayPal or something with a user giving you their user name and password and you crunching for them for x number of days for y cost. Alot of trust between people going on though would probably prevent some people from doing it.
I can see the OP's way of doing it though for challenges and other short term things. I could boost my crunching power for 2 weeks for example. |
|
|
|
When I saw this post, I hoped it was the reverse of what it actually is. I was hoping it was a way for someone else to use my home computer to run their projects for a small fee (which would have to be less than cloud providers to not be exploited). As someone who lives in an area with very low electricity cost, this would be attractive to me and then I would feel less bad about leaving my computer on 24/7 crunching things. Of course, you would also have to have the fee be associated with families of processors, and have requirements around stolen CPU time, etc.
You could do this on your own thru PayPal or something with a user giving you their user name and password and you crunching for them for x number of days for y cost. Alot of trust between people going on though would probably prevent some people from doing it.
I can see the OP's way of doing it though for challenges and other short term things. I could boost my crunching power for 2 weeks for example.
Which is why a coordinated marketplace would be much more attractive. When most of what you need is CPU power, why necessarily bother with a cloud provider when you can distribute the cloud over the world and create a true marketplace for compute resources? Potentially a project for me to put together once I retire.
____________
|
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2653 ID: 1 Credit: 109,635,410 RAC: 47,548
                     
|
Which is why a coordinated marketplace would be much more attractive. When most of what you need is CPU power, why necessarily bother with a cloud provider when you can distribute the cloud over the world and create a true marketplace for compute resources? Potentially a project for me to put together once I retire.
The goal of this project is to increase the computer time available to science. Moving your power to a different account for payment does not add to the total of the system, it essentially is buying credits without benefit to the science.
P.S. marketplaces like this already exist. |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2653 ID: 1 Credit: 109,635,410 RAC: 47,548
                     
|
An update regarding the tax: the tax situation has changed and I am no longer applying 21% VAT rate for participants from outside of the European Union / European Economic Area, as per Lithuanian law. This means that for a lot participants this has become significantly cheaper. |
|
|
|
I'm happy to support your service Rytis! It's quite convenient and I feel the costs are very good as well. You know how I feel as we spoke on Discord. You will get more business from me in the future. Very strong supporter of what you've put together here.
-Penguin |
|
|
tng Send message
Joined: 29 Aug 10 Posts: 499 ID: 66603 Credit: 50,812,070,359 RAC: 31,720,205
                                                    
|
Just got started, but seems very convenient and reasonably priced.
Would it be difficult to provide an option to to not keep a work queue?
____________
|
|
|
Ken_g6 Volunteer developer
 Send message
Joined: 4 Jul 06 Posts: 940 ID: 3110 Credit: 265,153,553 RAC: 110,745
                            
|
Would it be difficult to provide an option to to not keep a work queue?
You should be able to do that now, without involving the cloud setup. Go to your BOINC project preferences and set your "resource share" to 0. |
|
|
tng Send message
Joined: 29 Aug 10 Posts: 499 ID: 66603 Credit: 50,812,070,359 RAC: 31,720,205
                                                    
|
Would it be difficult to provide an option to to not keep a work queue?
You should be able to do that now, without involving the cloud setup. Go to your BOINC project preferences and set your "resource share" to 0.
Actually, it would seem that it's been too long since I looked at some of the options available from the Primegrid site. The "Computing preferences" link provides this. I had some settings defined in there, but it has been several years since I even looked at that (wish I had, could have used this).
____________
|
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2653 ID: 1 Credit: 109,635,410 RAC: 47,548
                     
|
Thanks for the good words!
Project preferences do apply, however, I fix some most common mistakes such as not running at 100% CPU or not using all of the available disk space using the preferences override files. |
|
|
|
When I tested it (for Rytis's money), I could find the hosts at /hosts_user.php, choose "Details", and then project them onto venues (Location) like Mercury and Venus. In that way they acquired different Computing preferences and PrimeGrid preferences, compared to my own hardware. (This, of course, will be well-known to a lot of people who participate with many computers.) /JeppeSN |
|
|
|
Thanks for creating this! It's a really easy way to get some extra computing running with very setup.
One thing I did want to report is that I'm seeing some fairly significant differences in task completion time between two seemingly identical instances. Both are 3-core Intel(R) Xeon(R) CPU E5-2630L v2 @ 2.40GHz machines, both running only PPSE tasks with multi-threading set to 1, both set to the same PG profile.
One tends to complete in ~1950sec: https://www.primegrid.com/results.php?hostid=981889
The other in ~1450sec: https://www.primegrid.com/results.php?hostid=981853
There's some variance over time, but generally one of them simply seems to be completing about 33% faster (and accordingly, has finished about 33% more tasks over the same period of time). Do you know if there would be some reason for this? I may just shut down the slower instance to make sure my credit is used more efficiently.
____________
|
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2653 ID: 1 Credit: 109,635,410 RAC: 47,548
                     
|
Sometimes there are "slow" instances that most likely run more demanding workloads from other customers on the same device. In case you encounter such an instance, I suggest to shut it down using the graceful shutdown option, and recreate it.
I will be adding monitoring so this process could get automated in the future. |
|
|
|
Sometimes there are "slow" instances that most likely run more demanding workloads from other customers on the same device. In case you encounter such an instance, I suggest to shut it down using the graceful shutdown option, and recreate it.
I will be adding monitoring so this process could get automated in the future.
Thanks, I'll try that out!
____________
|
|
|
|
Is it normal that now instance connect with PrimeGrid server once in every minute? |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2653 ID: 1 Credit: 109,635,410 RAC: 47,548
                     
|
Is it normal that now instance connect with PrimeGrid server once in every minute?
If it has low work, it will attempt to connect often. Right now it isn't a problem for the server, but obviously I'll fine tune the connection interval if it starts impacting in any way.
This has been implemented because of Rosetta@Home not having enough work and instances becoming idle. |
|
|